About the project

The Blue Nile floodplain in Sennar and Gezira is one of Sudan's most flood-exposed regions: towns, villages and the irrigated Gezira Scheme sit on flat clay plains a few metres above the river, and the record flood of August–September 2020 inundated farmland and homes along its banks. This dashboard maps where flooding occurred and where the ground is most prone to it. For any flood event it compares Sentinel-1 radar images from before and during the flood: radar sees through cloud, and calm water reflects the signal away from the satellite, so newly dark pixels in both VV and VH polarisations mark flood water. That observed extent is combined with the shape of the land (elevation, height above the nearest river and slope), the distance to the OpenStreetMap river network and recent rainfall into a flood risk map with four classes, from Low to Very High.

The whole analysis runs in your browser, with no server behind it. Elevation tiles, radar mosaics, rainfall records and river lines are fetched from public APIs (AWS Terrain Tiles, Microsoft Planetary Computer or Sentinel Hub, Open-Meteo and Overpass), decoded pixel by pixel, and processed on a ~300 m grid: slope from the elevation model, an exact distance transform to the rivers, threshold-based water detection on the radar, and a weighted risk score for every cell. The model is deliberately transparent. You can change the flood dates, the radar water thresholds and the weight of each factor, then press Recompute Risk to see how the map responds, and click anywhere to inspect the values behind a cell. It is a screening tool for exploring flood exposure rather than a calibrated hazard map, but it shows how open satellite data and client-side GeoAI can deliver rapid flood insight for data-scarce regions.